Don't go into federation catch up mode so easily #9561

erikjohnston · 2021-03-08T17:59:06Z

Federation catch up mode is very inefficient if the number of events
that the remote server has missed is small, since handling gaps can be
very expensive, c.f. #9492.

Instead of going into catch up mode whenever we see an error, we instead
do so only if we've backed off from trying the remote for more than an
hour (the assumption being that in such a case it is more than a
transient failure).

Federation catch up mode is very inefficient if the number of events that the remote server has missed is small, since handling gaps can be very expensive, c.f. #9492. Instead of going into catch up mode whenever we see an error, we instead do so only if we've backed off from trying the remote for more than an hour (the assumption being that in such a case it is more than a transient failure).

anoadragon453

So in the case of someone who's gone offline for 55 minutes, they'll still get missed messages as transactions will still be retried from other servers after only that short amount of downtime?

anoadragon453 · 2021-03-09T11:35:00Z

synapse/federation/sender/per_destination_queue.py

            for p in pending_pdus:
                logger.info(
                    "Failed to send event %s to %s", p.event_id, self._destination
                )


Unrelated, but I wonder if we should try to log the event IDs on one line here instead of 50. In another PR though.

erikjohnston · 2021-03-09T11:43:35Z

So in the case of someone who's gone offline for 55 minutes, they'll still get missed messages as transactions will still be retried from other servers after only that short amount of downtime?

That's the idea.

But I think I've messed it up....

anoadragon453 · 2021-03-09T11:58:16Z

I assume #9567 is intended to sort out the usecase I mentioned? In which case this lgtm.

erikjohnston · 2021-03-10T17:11:48Z

For context there was a problem where we dropped the PDUs that were included in a transaction even if it failed, which meant if we didn't go into catchup mode those events would not be sent out. The above commits make it so that we requeue up events if we fail to send them (plus a bunch of refactors to make that easier).

This isn't quite ready for review yet though.

erikjohnston · 2021-03-11T11:52:51Z

Ok, I think this is now ready for review again. Probably easiest to look at it commit-by-commit

clokep

I think that this looks good overall, but I left some comments. Most of them can probably be ignored / deferred to another PR, but a couple are trying to understand what this is doing. 😄

clokep · 2021-03-12T14:41:22Z

synapse/federation/sender/transaction_manager.py

-        Returns:
-            True iff the transaction was successful


Is it worth mentioning what this raises while we're here? 😄

synapse/federation/sender/per_destination_queue.py

clokep · 2021-03-12T15:11:48Z

synapse/federation/sender/per_destination_queue.py

+        pending_presence = self.queue._pending_presence
+        self.queue._pending_presence = {}


This will cause us to declare a new dict in some cases when we don't need to, it seems we could (instead) check if self.queue._pending_presence below and only overwrite this value if necessary (i.e. the pending_presence variable is unnecessary).

clokep · 2021-03-12T15:14:22Z

synapse/federation/sender/per_destination_queue.py

+            len(pending_edus) < MAX_EDUS_PER_TRANSACTION
+            and self.queue._pending_edus_keyed
+        ):
+            _, val = self.queue._pending_edus_keyed.popitem()


This is copied & pasted, but popitem is LIFO so it is theoretically possible you'll never reach the bottom of this queue, which is annoying. 😢

Hmm, yeah. In a separate PR i was going to make it so that we don't drop the EDUs if the transaction fails (in the same way as for PDUs), so it'll probably get magically fixed there?

synapse/federation/sender/per_destination_queue.py

clokep

LGTM!

Synapse 1.30.0rc1 (2021-03-16) ============================== Note that this release deprecates the ability for appservices to call `POST /_matrix/client/r0/register` without the body parameter `type`. Appservice developers should use a `type` value of `m.login.application_service` as per [the spec](https://matrix.org/docs/spec/application_service/r0.1.2#server-admin-style-permissions). In future releases, calling this endpoint with an access token - but without a `m.login.application_service` type - will fail. Features -------- - Add prometheus metrics for number of users successfully registering and logging in. ([\#9510](#9510), [\#9511](#9511), [\#9573](#9573)) - Add `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time` prometheus metrics, which monitor federation delays by reporting the timestamps of messages sent and received to a set of remote servers. ([\#9540](#9540)) - Add support for generating JSON Web Tokens dynamically for use as OIDC client secrets. ([\#9549](#9549)) - Optimise handling of incomplete room history for incoming federation. ([\#9601](#9601)) - Finalise support for allowing clients to pick an SSO Identity Provider ([MSC2858](matrix-org/matrix-spec-proposals#2858)). ([\#9617](#9617)) - Tell spam checker modules about the SSO IdP a user registered through if one was used. ([\#9626](#9626)) Bugfixes -------- - Fix long-standing bug when generating thumbnails for some images with transparency: `TypeError: cannot unpack non-iterable int object`. ([\#9473](#9473)) - Purge chain cover indexes for events that were purged prior to Synapse v1.29.0. ([\#9542](#9542), [\#9583](#9583)) - Fix bug where federation requests were not correctly retried on 5xx responses. ([\#9567](#9567)) - Fix re-activating an account via the admin API when local passwords are disabled. ([\#9587](#9587)) - Fix a bug introduced in Synapse 1.20 which caused incoming federation transactions to stack up, causing slow recovery from outages. ([\#9597](#9597)) - Fix a bug introduced in v1.28.0 where the OpenID Connect callback endpoint could error with a `MacaroonInitException`. ([\#9620](#9620)) - Fix Internal Server Error on `GET /_synapse/client/saml2/authn_response` request. ([\#9623](#9623)) Updates to the Docker image --------------------------- - Use jemalloc if available in docker. ([\#8553](#8553)) Improved Documentation ---------------------- - Add relayd entry to reverse proxy example configurations. ([\#9508](#9508)) - Improve the SAML2 upgrade notes for 1.27.0. ([\#9550](#9550)) - Link to the "List user's media" admin API from the media admin API docs. ([\#9571](#9571)) - Clarify the spam checker modules documentation example to mention that `parse_config` is a required method. ([\#9580](#9580)) - Clarify the sample configuration for `stats` settings. ([\#9604](#9604)) Deprecations and Removals ------------------------- - The `synapse_federation_last_sent_pdu_age` and `synapse_federation_last_received_pdu_age` prometheus metrics have been removed. They are replaced by `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time`. ([\#9540](#9540)) - Registering an Application Service user without using the `m.login.application_service` login type will be unsupported in an upcoming Synapse release. ([\#9559](#9559)) Internal Changes ---------------- - Add tests to ResponseCache. ([\#9458](#9458)) - Add type hints to purge room and server notice admin API. ([\#9520](#9520)) - Add extra logging to ObservableDeferred when callbacks throw exceptions. ([\#9523](#9523)) - Fix incorrect type hints. ([\#9528](#9528), [\#9543](#9543), [\#9591](#9591), [\#9608](#9608), [\#9618](#9618)) - Add an additional test for purging a room. ([\#9541](#9541)) - Add a `.git-blame-ignore-revs` file with the hashes of auto-formatting. ([\#9560](#9560)) - Increase the threshold before which outbound federation to a server goes into "catch up" mode, which is expensive for the remote server to handle. ([\#9561](#9561)) - Fix spurious errors reported by the `config-lint.sh` script. ([\#9562](#9562)) - Fix type hints and tests for BlacklistingAgentWrapper and BlacklistingReactorWrapper. ([\#9563](#9563)) - Do not have mypy ignore type hints from unpaddedbase64. ([\#9568](#9568)) - Improve efficiency of calculating the auth chain in large rooms. ([\#9576](#9576)) - Convert `synapse.types.Requester` to an `attrs` class. ([\#9586](#9586)) - Add logging for redis connection setup. ([\#9590](#9590)) - Improve logging when processing incoming transactions. ([\#9596](#9596)) - Remove unused `stats.retention` setting, and emit a warning if stats are disabled. ([\#9604](#9604)) - Prevent attempting to bundle aggregations for state events in /context APIs. ([\#9619](#9619))

Synapse 1.30.0 (2021-03-22) =========================== Note that this release deprecates the ability for appservices to call `POST /_matrix/client/r0/register` without the body parameter `type`. Appservice developers should use a `type` value of `m.login.application_service` as per [the spec](https://matrix.org/docs/spec/application_service/r0.1.2#server-admin-style-permissions). In future releases, calling this endpoint with an access token - but without a `m.login.application_service` type - will fail. No significant changes. Synapse 1.30.0rc1 (2021-03-16) ============================== Features -------- - Add prometheus metrics for number of users successfully registering and logging in. ([\#9510](#9510), [\#9511](#9511), [\#9573](#9573)) - Add `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time` prometheus metrics, which monitor federation delays by reporting the timestamps of messages sent and received to a set of remote servers. ([\#9540](#9540)) - Add support for generating JSON Web Tokens dynamically for use as OIDC client secrets. ([\#9549](#9549)) - Optimise handling of incomplete room history for incoming federation. ([\#9601](#9601)) - Finalise support for allowing clients to pick an SSO Identity Provider ([MSC2858](matrix-org/matrix-spec-proposals#2858)). ([\#9617](#9617)) - Tell spam checker modules about the SSO IdP a user registered through if one was used. ([\#9626](#9626)) Bugfixes -------- - Fix long-standing bug when generating thumbnails for some images with transparency: `TypeError: cannot unpack non-iterable int object`. ([\#9473](#9473)) - Purge chain cover indexes for events that were purged prior to Synapse v1.29.0. ([\#9542](#9542), [\#9583](#9583)) - Fix bug where federation requests were not correctly retried on 5xx responses. ([\#9567](#9567)) - Fix re-activating an account via the admin API when local passwords are disabled. ([\#9587](#9587)) - Fix a bug introduced in Synapse 1.20 which caused incoming federation transactions to stack up, causing slow recovery from outages. ([\#9597](#9597)) - Fix a bug introduced in v1.28.0 where the OpenID Connect callback endpoint could error with a `MacaroonInitException`. ([\#9620](#9620)) - Fix Internal Server Error on `GET /_synapse/client/saml2/authn_response` request. ([\#9623](#9623)) Updates to the Docker image --------------------------- - Make use of an improved malloc implementation (`jemalloc`) in the docker image. ([\#8553](#8553)) Improved Documentation ---------------------- - Add relayd entry to reverse proxy example configurations. ([\#9508](#9508)) - Improve the SAML2 upgrade notes for 1.27.0. ([\#9550](#9550)) - Link to the "List user's media" admin API from the media admin API docs. ([\#9571](#9571)) - Clarify the spam checker modules documentation example to mention that `parse_config` is a required method. ([\#9580](#9580)) - Clarify the sample configuration for `stats` settings. ([\#9604](#9604)) Deprecations and Removals ------------------------- - The `synapse_federation_last_sent_pdu_age` and `synapse_federation_last_received_pdu_age` prometheus metrics have been removed. They are replaced by `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time`. ([\#9540](#9540)) - Registering an Application Service user without using the `m.login.application_service` login type will be unsupported in an upcoming Synapse release. ([\#9559](#9559)) Internal Changes ---------------- - Add tests to ResponseCache. ([\#9458](#9458)) - Add type hints to purge room and server notice admin API. ([\#9520](#9520)) - Add extra logging to ObservableDeferred when callbacks throw exceptions. ([\#9523](#9523)) - Fix incorrect type hints. ([\#9528](#9528), [\#9543](#9543), [\#9591](#9591), [\#9608](#9608), [\#9618](#9618)) - Add an additional test for purging a room. ([\#9541](#9541)) - Add a `.git-blame-ignore-revs` file with the hashes of auto-formatting. ([\#9560](#9560)) - Increase the threshold before which outbound federation to a server goes into "catch up" mode, which is expensive for the remote server to handle. ([\#9561](#9561)) - Fix spurious errors reported by the `config-lint.sh` script. ([\#9562](#9562)) - Fix type hints and tests for BlacklistingAgentWrapper and BlacklistingReactorWrapper. ([\#9563](#9563)) - Do not have mypy ignore type hints from unpaddedbase64. ([\#9568](#9568)) - Improve efficiency of calculating the auth chain in large rooms. ([\#9576](#9576)) - Convert `synapse.types.Requester` to an `attrs` class. ([\#9586](#9586)) - Add logging for redis connection setup. ([\#9590](#9590)) - Improve logging when processing incoming transactions. ([\#9596](#9596)) - Remove unused `stats.retention` setting, and emit a warning if stats are disabled. ([\#9604](#9604)) - Prevent attempting to bundle aggregations for state events in /context APIs. ([\#9619](#9619))

@dklimpel

Synapse 1.32.2 (2021-04-22) =========================== This release includes a fix for a regression introduced in 1.32.0. Bugfixes -------- - Fix a regression in Synapse 1.32.0 and 1.32.1 which caused `LoggingContext` errors in plugins. ([\#9857](matrix-org/synapse#9857)) Synapse 1.32.1 (2021-04-21) =========================== This release fixes [a regression](matrix-org/synapse#9853) in Synapse 1.32.0 that caused connected Prometheus instances to become unstable. However, as this release is still subject to the `LoggingContext` change in 1.32.0, it is recommended to remain on or downgrade to 1.31.0. Bugfixes -------- - Fix a regression in Synapse 1.32.0 which caused Synapse to report large numbers of Prometheus time series, potentially overwhelming Prometheus instances. ([\#9854](matrix-org/synapse#9854)) Synapse 1.32.0 (2021-04-20) =========================== **Note:** This release introduces [a regression](matrix-org/synapse#9853) that can overwhelm connected Prometheus instances. This issue was not present in 1.32.0rc1. If affected, it is recommended to downgrade to 1.31.0 in the meantime, and follow [these instructions](matrix-org/synapse#9854 (comment)) to clean up any excess writeahead logs. **Note:** This release also mistakenly included a change that may affected Synapse modules that import `synapse.logging.context.LoggingContext`, such as [synapse-s3-storage-provider](https://github.com/matrix-org/synapse-s3-storage-provider). This will be fixed in a later Synapse version. **Note:** This release requires Python 3.6+ and Postgres 9.6+ or SQLite 3.22+. This release removes the deprecated `GET /_synapse/admin/v1/users/<user_id>` admin API. Please use the [v2 API](https://github.com/matrix-org/synapse/blob/develop/docs/admin_api/user_admin_api.rst#query-user-account) instead, which has improved capabilities. This release requires Application Services to use type `m.login.application_service` when registering users via the `/_matrix/client/r0/register` endpoint to comply with the spec. Please ensure your Application Services are up to date. If you are using the `packages.matrix.org` Debian repository for Synapse packages, note that we have recently updated the expiry date on the gpg signing key. If you see an error similar to `The following signatures were invalid: EXPKEYSIG F473DD4473365DE1`, you will need to get a fresh copy of the keys. You can do so with: ```sh sudo wget -O /usr/share/keyrings/matrix-org-archive-keyring.gpg https://packages.matrix.org/debian/matrix-org-archive-keyring.gpg ``` Bugfixes -------- - Fix the log lines of nested logging contexts. Broke in 1.32.0rc1. ([\#9829](matrix-org/synapse#9829)) Synapse 1.32.0rc1 (2021-04-13) ============================== Features -------- - Add a Synapse module for routing presence updates between users. ([\#9491](matrix-org/synapse#9491)) - Add an admin API to manage ratelimit for a specific user. ([\#9648](matrix-org/synapse#9648)) - Include request information in structured logging output. ([\#9654](matrix-org/synapse#9654)) - Add `order_by` to the admin API `GET /_synapse/admin/v2/users`. Contributed by @dklimpel. ([\#9691](matrix-org/synapse#9691)) - Replace the `room_invite_state_types` configuration setting with `room_prejoin_state`. ([\#9700](matrix-org/synapse#9700)) - Add experimental support for [MSC3083](matrix-org/matrix-spec-proposals#3083): restricting room access via group membership. ([\#9717](matrix-org/synapse#9717), [\#9735](matrix-org/synapse#9735)) - Update experimental support for Spaces: include `m.room.create` in the room state sent with room-invites. ([\#9710](matrix-org/synapse#9710)) - Synapse now requires Python 3.6 or later. It also requires Postgres 9.6 or later or SQLite 3.22 or later. ([\#9766](matrix-org/synapse#9766)) Bugfixes -------- - Prevent `synapse_forward_extremities` and `synapse_excess_extremity_events` Prometheus metrics from initially reporting zero-values after startup. ([\#8926](matrix-org/synapse#8926)) - Fix recently added ratelimits to correctly honour the application service `rate_limited` flag. ([\#9711](matrix-org/synapse#9711)) - Fix longstanding bug which caused `duplicate key value violates unique constraint "remote_media_cache_thumbnails_media_origin_media_id_thumbna_key"` errors. ([\#9725](matrix-org/synapse#9725)) - Fix bug where sharded federation senders could get stuck repeatedly querying the DB in a loop, using lots of CPU. ([\#9770](matrix-org/synapse#9770)) - Fix duplicate logging of exceptions thrown during federation transaction processing. ([\#9780](matrix-org/synapse#9780)) Updates to the Docker image --------------------------- - Move opencontainers labels to the final Docker image such that users can inspect them. ([\#9765](matrix-org/synapse#9765)) Improved Documentation ---------------------- - Make the `allowed_local_3pids` regex example in the sample config stricter. ([\#9719](matrix-org/synapse#9719)) Deprecations and Removals ------------------------- - Remove old admin API `GET /_synapse/admin/v1/users/<user_id>`. ([\#9401](matrix-org/synapse#9401)) - Make `/_matrix/client/r0/register` expect a type of `m.login.application_service` when an Application Service registers a user, to align with [the relevant spec](https://spec.matrix.org/unstable/application-service-api/#server-admin-style-permissions). ([\#9548](matrix-org/synapse#9548)) Internal Changes ---------------- - Replace deprecated `imp` module with successor `importlib`. Contributed by Cristina Muñoz. ([\#9718](matrix-org/synapse#9718)) - Experiment with GitHub Actions for CI. ([\#9661](matrix-org/synapse#9661)) - Introduce flake8-bugbear to the test suite and fix some of its lint violations. ([\#9682](matrix-org/synapse#9682)) - Update `scripts-dev/complement.sh` to use a local checkout of Complement, allow running a subset of tests and have it use Synapse's Complement test blacklist. ([\#9685](matrix-org/synapse#9685)) - Improve Jaeger tracing for `to_device` messages. ([\#9686](matrix-org/synapse#9686)) - Add release helper script for automating part of the Synapse release process. ([\#9713](matrix-org/synapse#9713)) - Add type hints to expiring cache. ([\#9730](matrix-org/synapse#9730)) - Convert various testcases to `HomeserverTestCase`. ([\#9736](matrix-org/synapse#9736)) - Start linting mypy with `no_implicit_optional`. ([\#9742](matrix-org/synapse#9742)) - Add missing type hints to federation handler and server. ([\#9743](matrix-org/synapse#9743)) - Check that a `ConfigError` is raised, rather than simply `Exception`, when appropriate in homeserver config file generation tests. ([\#9753](matrix-org/synapse#9753)) - Fix incompatibility with `tox` 2.5. ([\#9769](matrix-org/synapse#9769)) - Enable Complement tests for [MSC2946](matrix-org/matrix-spec-proposals#2946): Spaces Summary API. ([\#9771](matrix-org/synapse#9771)) - Use mock from the standard library instead of a separate package. ([\#9772](matrix-org/synapse#9772)) - Update Black configuration to target Python 3.6. ([\#9781](matrix-org/synapse#9781)) - Add option to skip unit tests when building Debian packages. ([\#9793](matrix-org/synapse#9793)) Synapse 1.31.0 (2021-04-06) =========================== **Note:** As announced in v1.25.0, and in line with the deprecation policy for platform dependencies, this is the last release to support Python 3.5 and PostgreSQL 9.5. Future versions of Synapse will require Python 3.6+ and PostgreSQL 9.6+, as per our [deprecation policy](docs/deprecation_policy.md). This is also the last release that the Synapse team will be publishing packages for Debian Stretch and Ubuntu Xenial. Improved Documentation ---------------------- - Add a document describing the deprecation policy for platform dependencies. ([\#9723](matrix-org/synapse#9723)) Internal Changes ---------------- - Revert using `dmypy run` in lint script. ([\#9720](matrix-org/synapse#9720)) - Pin flake8-bugbear's version. ([\#9734](matrix-org/synapse#9734)) Synapse 1.31.0rc1 (2021-03-30) ============================== Features -------- - Add support to OpenID Connect login for requiring attributes on the `userinfo` response. Contributed by Hubbe King. ([\#9609](matrix-org/synapse#9609)) - Add initial experimental support for a "space summary" API. ([\#9643](matrix-org/synapse#9643), [\#9652](matrix-org/synapse#9652), [\#9653](matrix-org/synapse#9653)) - Add support for the busy presence state as described in [MSC3026](matrix-org/matrix-spec-proposals#3026). ([\#9644](matrix-org/synapse#9644)) - Add support for credentials for proxy authentication in the `HTTPS_PROXY` environment variable. ([\#9657](matrix-org/synapse#9657)) Bugfixes -------- - Fix a longstanding bug that could cause issues when editing a reply to a message. ([\#9585](matrix-org/synapse#9585)) - Fix the `/capabilities` endpoint to return `m.change_password` as disabled if the local password database is not used for authentication. Contributed by @dklimpel. ([\#9588](matrix-org/synapse#9588)) - Check if local passwords are enabled before setting them for the user. ([\#9636](matrix-org/synapse#9636)) - Fix a bug where federation sending can stall due to `concurrent access` database exceptions when it falls behind. ([\#9639](matrix-org/synapse#9639)) - Fix a bug introduced in Synapse 1.30.1 which meant the suggested `pip` incantation to install an updated `cryptography` was incorrect. ([\#9699](matrix-org/synapse#9699)) Updates to the Docker image --------------------------- - Speed up Docker builds and make it nicer to test against Complement while developing (install all dependencies before copying the project). ([\#9610](matrix-org/synapse#9610)) - Include [opencontainers labels](https://github.com/opencontainers/image-spec/blob/master/annotations.md#pre-defined-annotation-keys) in the Docker image. ([\#9612](matrix-org/synapse#9612)) Improved Documentation ---------------------- - Clarify that `register_new_matrix_user` is present also when installed via non-pip package. ([\#9074](matrix-org/synapse#9074)) - Update source install documentation to mention platform prerequisites before the source install steps. ([\#9667](matrix-org/synapse#9667)) - Improve worker documentation for fallback/web auth endpoints. ([\#9679](matrix-org/synapse#9679)) - Update the sample configuration for OIDC authentication. ([\#9695](matrix-org/synapse#9695)) Internal Changes ---------------- - Preparatory steps for removing redundant `outlier` data from `event_json.internal_metadata` column. ([\#9411](matrix-org/synapse#9411)) - Add type hints to the caching module. ([\#9442](matrix-org/synapse#9442)) - Introduce flake8-bugbear to the test suite and fix some of its lint violations. ([\#9499](matrix-org/synapse#9499), [\#9659](matrix-org/synapse#9659)) - Add additional type hints to the Homeserver object. ([\#9631](matrix-org/synapse#9631), [\#9638](matrix-org/synapse#9638), [\#9675](matrix-org/synapse#9675), [\#9681](matrix-org/synapse#9681)) - Only save remote cross-signing and device keys if they're different from the current ones. ([\#9634](matrix-org/synapse#9634)) - Rename storage function to fix spelling and not conflict with another function's name. ([\#9637](matrix-org/synapse#9637)) - Improve performance of federation catch up by sending the latest events in the room to the remote, rather than just the last event sent by the local server. ([\#9640](matrix-org/synapse#9640), [\#9664](matrix-org/synapse#9664)) - In the `federation_client` commandline client, stop automatically adding the URL prefix, so that servlets on other prefixes can be tested. ([\#9645](matrix-org/synapse#9645)) - In the `federation_client` commandline client, handle inline `signing_key`s in `homeserver.yaml`. ([\#9647](matrix-org/synapse#9647)) - Fixed some antipattern issues to improve code quality. ([\#9649](matrix-org/synapse#9649)) - Add a storage method for pulling all current user presence state from the database. ([\#9650](matrix-org/synapse#9650)) - Import `HomeServer` from the proper module. ([\#9665](matrix-org/synapse#9665)) - Increase default join ratelimiting burst rate. ([\#9674](matrix-org/synapse#9674)) - Add type hints to third party event rules and visibility modules. ([\#9676](matrix-org/synapse#9676)) - Bump mypy-zope to 0.2.13 to fix "Cannot determine consistent method resolution order (MRO)" errors when running mypy a second time. ([\#9678](matrix-org/synapse#9678)) - Use interpreter from `$PATH` via `/usr/bin/env` instead of absolute paths in various scripts. ([\#9689](matrix-org/synapse#9689)) - Make it possible to use `dmypy`. ([\#9692](matrix-org/synapse#9692)) - Suppress "CryptographyDeprecationWarning: int_from_bytes is deprecated". ([\#9698](matrix-org/synapse#9698)) - Use `dmypy run` in lint script for improved performance in type-checking while developing. ([\#9701](matrix-org/synapse#9701)) - Fix undetected mypy error when using Python 3.6. ([\#9703](matrix-org/synapse#9703)) - Fix type-checking CI on develop. ([\#9709](matrix-org/synapse#9709)) Synapse 1.30.1 (2021-03-26) =========================== This release is identical to Synapse 1.30.0, with the exception of explicitly setting a minimum version of Python's Cryptography library to ensure that users of Synapse are protected from the recent [OpenSSL security advisories](https://mta.openssl.org/pipermail/openssl-announce/2021-March/000198.html), especially CVE-2021-3449. Note that Cryptography defaults to bundling its own statically linked copy of OpenSSL, which means that you may not be protected by your operating system's security updates. It's also worth noting that Cryptography no longer supports Python 3.5, so admins deploying to older environments may not be protected against this or future vulnerabilities. Synapse will be dropping support for Python 3.5 at the end of March. Updates to the Docker image --------------------------- - Ensure that the docker container has up to date versions of openssl. ([\#9697](matrix-org/synapse#9697)) Internal Changes ---------------- - Enforce that `cryptography` dependency is up to date to ensure it has the most recent openssl patches. ([\#9697](matrix-org/synapse#9697)) Synapse 1.30.0 (2021-03-22) =========================== Note that this release deprecates the ability for appservices to call `POST /_matrix/client/r0/register` without the body parameter `type`. Appservice developers should use a `type` value of `m.login.application_service` as per [the spec](https://matrix.org/docs/spec/application_service/r0.1.2#server-admin-style-permissions). In future releases, calling this endpoint with an access token - but without a `m.login.application_service` type - will fail. No significant changes. Synapse 1.30.0rc1 (2021-03-16) ============================== Features -------- - Add prometheus metrics for number of users successfully registering and logging in. ([\#9510](matrix-org/synapse#9510), [\#9511](matrix-org/synapse#9511), [\#9573](matrix-org/synapse#9573)) - Add `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time` prometheus metrics, which monitor federation delays by reporting the timestamps of messages sent and received to a set of remote servers. ([\#9540](matrix-org/synapse#9540)) - Add support for generating JSON Web Tokens dynamically for use as OIDC client secrets. ([\#9549](matrix-org/synapse#9549)) - Optimise handling of incomplete room history for incoming federation. ([\#9601](matrix-org/synapse#9601)) - Finalise support for allowing clients to pick an SSO Identity Provider ([MSC2858](matrix-org/matrix-spec-proposals#2858)). ([\#9617](matrix-org/synapse#9617)) - Tell spam checker modules about the SSO IdP a user registered through if one was used. ([\#9626](matrix-org/synapse#9626)) Bugfixes -------- - Fix long-standing bug when generating thumbnails for some images with transparency: `TypeError: cannot unpack non-iterable int object`. ([\#9473](matrix-org/synapse#9473)) - Purge chain cover indexes for events that were purged prior to Synapse v1.29.0. ([\#9542](matrix-org/synapse#9542), [\#9583](matrix-org/synapse#9583)) - Fix bug where federation requests were not correctly retried on 5xx responses. ([\#9567](matrix-org/synapse#9567)) - Fix re-activating an account via the admin API when local passwords are disabled. ([\#9587](matrix-org/synapse#9587)) - Fix a bug introduced in Synapse 1.20 which caused incoming federation transactions to stack up, causing slow recovery from outages. ([\#9597](matrix-org/synapse#9597)) - Fix a bug introduced in v1.28.0 where the OpenID Connect callback endpoint could error with a `MacaroonInitException`. ([\#9620](matrix-org/synapse#9620)) - Fix Internal Server Error on `GET /_synapse/client/saml2/authn_response` request. ([\#9623](matrix-org/synapse#9623)) Updates to the Docker image --------------------------- - Make use of an improved malloc implementation (`jemalloc`) in the docker image. ([\#8553](matrix-org/synapse#8553)) Improved Documentation ---------------------- - Add relayd entry to reverse proxy example configurations. ([\#9508](matrix-org/synapse#9508)) - Improve the SAML2 upgrade notes for 1.27.0. ([\#9550](matrix-org/synapse#9550)) - Link to the "List user's media" admin API from the media admin API docs. ([\#9571](matrix-org/synapse#9571)) - Clarify the spam checker modules documentation example to mention that `parse_config` is a required method. ([\#9580](matrix-org/synapse#9580)) - Clarify the sample configuration for `stats` settings. ([\#9604](matrix-org/synapse#9604)) Deprecations and Removals ------------------------- - The `synapse_federation_last_sent_pdu_age` and `synapse_federation_last_received_pdu_age` prometheus metrics have been removed. They are replaced by `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time`. ([\#9540](matrix-org/synapse#9540)) - Registering an Application Service user without using the `m.login.application_service` login type will be unsupported in an upcoming Synapse release. ([\#9559](matrix-org/synapse#9559)) Internal Changes ---------------- - Add tests to ResponseCache. ([\#9458](matrix-org/synapse#9458)) - Add type hints to purge room and server notice admin API. ([\#9520](matrix-org/synapse#9520)) - Add extra logging to ObservableDeferred when callbacks throw exceptions. ([\#9523](matrix-org/synapse#9523)) - Fix incorrect type hints. ([\#9528](matrix-org/synapse#9528), [\#9543](matrix-org/synapse#9543), [\#9591](matrix-org/synapse#9591), [\#9608](matrix-org/synapse#9608), [\#9618](matrix-org/synapse#9618)) - Add an additional test for purging a room. ([\#9541](matrix-org/synapse#9541)) - Add a `.git-blame-ignore-revs` file with the hashes of auto-formatting. ([\#9560](matrix-org/synapse#9560)) - Increase the threshold before which outbound federation to a server goes into "catch up" mode, which is expensive for the remote server to handle. ([\#9561](matrix-org/synapse#9561)) - Fix spurious errors reported by the `config-lint.sh` script. ([\#9562](matrix-org/synapse#9562)) - Fix type hints and tests for BlacklistingAgentWrapper and BlacklistingReactorWrapper. ([\#9563](matrix-org/synapse#9563)) - Do not have mypy ignore type hints from unpaddedbase64. ([\#9568](matrix-org/synapse#9568)) - Improve efficiency of calculating the auth chain in large rooms. ([\#9576](matrix-org/synapse#9576)) - Convert `synapse.types.Requester` to an `attrs` class. ([\#9586](matrix-org/synapse#9586)) - Add logging for redis connection setup. ([\#9590](matrix-org/synapse#9590)) - Improve logging when processing incoming transactions. ([\#9596](matrix-org/synapse#9596)) - Remove unused `stats.retention` setting, and emit a warning if stats are disabled. ([\#9604](matrix-org/synapse#9604)) - Prevent attempting to bundle aggregations for state events in /context APIs. ([\#9619](matrix-org/synapse#9619))

erikjohnston added 2 commits March 8, 2021 17:54

Newsfile

52510c5

erikjohnston requested a review from a team March 8, 2021 18:17

anoadragon453 reviewed Mar 9, 2021

View reviewed changes

erikjohnston added 4 commits March 10, 2021 11:43

Name parameters in function call for clarity

7becd94

Use a flag to mark if there is new stuff, rather than an atomic section

454da8a

Have send_new_transaction always raise on error

4f04ecc

Move pulling from queues to a async context manager

d6effcf

erikjohnston force-pushed the erikj/less_catchup branch from 6b2ef9c to e1c992a Compare March 10, 2021 17:04

erikjohnston force-pushed the erikj/less_catchup branch from e1c992a to 5aca0b9 Compare March 10, 2021 17:13

Requeue events for sending if transaction fails

d0bf228

erikjohnston force-pushed the erikj/less_catchup branch from 5aca0b9 to d0bf228 Compare March 10, 2021 17:33

Fix tests

b90c517

erikjohnston requested a review from a team March 11, 2021 11:53

clokep reviewed Mar 12, 2021

View reviewed changes

Review comments

216d12d

erikjohnston requested a review from clokep March 15, 2021 14:13

clokep approved these changes Mar 15, 2021

View reviewed changes

erikjohnston merged commit 026503f into develop Mar 15, 2021

erikjohnston deleted the erikj/less_catchup branch March 15, 2021 14:42

ShadowJonathan mentioned this pull request Mar 16, 2021

Do federation catchup in "buckets" #9632

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Don't go into federation catch up mode so easily #9561

Don't go into federation catch up mode so easily #9561

erikjohnston commented Mar 8, 2021

anoadragon453 left a comment

anoadragon453 Mar 9, 2021

erikjohnston commented Mar 9, 2021

anoadragon453 commented Mar 9, 2021

erikjohnston commented Mar 10, 2021

erikjohnston commented Mar 11, 2021

clokep left a comment

clokep Mar 12, 2021

clokep Mar 12, 2021

clokep Mar 12, 2021

erikjohnston Mar 15, 2021

clokep left a comment

		pending_presence = self.queue._pending_presence
		self.queue._pending_presence = {}

Don't go into federation catch up mode so easily #9561

Don't go into federation catch up mode so easily #9561

Conversation

erikjohnston commented Mar 8, 2021

anoadragon453 left a comment

Choose a reason for hiding this comment

anoadragon453 Mar 9, 2021

Choose a reason for hiding this comment

erikjohnston commented Mar 9, 2021

anoadragon453 commented Mar 9, 2021

erikjohnston commented Mar 10, 2021

erikjohnston commented Mar 11, 2021

clokep left a comment

Choose a reason for hiding this comment

clokep Mar 12, 2021

Choose a reason for hiding this comment

clokep Mar 12, 2021

Choose a reason for hiding this comment

clokep Mar 12, 2021

Choose a reason for hiding this comment

erikjohnston Mar 15, 2021

Choose a reason for hiding this comment

clokep left a comment

Choose a reason for hiding this comment